OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Platonictransformers

Platonictransformers is an open-source Python research repository implementing Transformer architectures with geometric inductive biases and approximate equivariance. Its code includes frame-relative attention, symmetry-group components, equivariant scalar and vector input/output handling, and a linear-time dynamic group-convolution variant. Dataset loaders and a unified entry point support documented research workfl

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Platonictransformers — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a research task involves geometric structure and you want code and example workflows for investigating equivariant Transformer designs.

Who it is for

  • Researchers exploring geometric deep learning and group-equivariant neural networks.
  • Developers prototyping Transformer methods for molecular, materials, image or 3D learning tasks.

Documented uses

  • Add geometric inductive biases to a Transformer architecture.
  • Explore group-equivariant attention and dynamic group convolution.
  • Run documented benchmark training workflows for CIFAR-10, QM9 or OMol.
Before you choose

Requirements & limits

The public repository is identified as open source and lists an MIT license.

  • The stated minimums are Python 3.12+ and PyTorch 2.3+; CUDA 12.1+ is specified for GPU support.
  • FlashAttention-2 is optional and recommended for the default fast attention configurations.
  • OMol training has additional documented dependencies, including FairChem; Lightning may require installation from source in some environments.
  • The README’s benchmark examples use CIFAR-10, QM9 and OMol, while repository structure also lists ImageNet and ScanObjectNN training scripts.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the documented research environment

Create and activate a Python 3.12 environment and install the basic build tools before following the project’s experiment-specific dependency instructions.

Before you start

  • Python 3.12 or later
  • The repository README recommends uv or plain pip; uv is not required.

Steps

  1. Create a virtual environment with the command below.

    Documented command
    uv venv --python 3.12 venv

  2. Activate it with source venv/bin/activate.

  3. Upgrade pip, wheel and setuptools using the README’s command.

Expected outcome

A Python 3.12 virtual environment with updated packaging tools, ready for the next documented installation stage.

Guide limitations

  • These steps cover environment setup only, not installation of all project dependencies or running a training job.
  • CUDA 12.1 or later is listed for GPU support; PyTorch 2.3 or later is a stated prerequisite.
  • The README notes that Lightning may require its documented source-install route in some environments.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
niazoys/PlatonicTransformers · Official homepage title · retained in its source language
How we review sourcesSuggest a correction